| An isothermal-flow, two-control-volume, bulk-flow rotordynamic analysis for smooth-rotor/honeycomb-stator gas annular seals is developed with two different viscous shear stress models. One model employs the directionally homogeneous viscous shear stress model of Hirs (1973) and is anchored to friction-factors extracted from pressure-driven flow measurements between flat plats, Ha et al. (1992). The other (New) model treats the viscous shear stress as being non-directionally homogeneous and is anchored to the same friction-factors from flat plat tests and others extracted from drag-torque measurements from a newly built test apparatus.;The two-control-volume bulk-flow analysis demonstrates that the cells of honeycomb act to reduce the effective acoustic velocity within the seal and can drop seal acoustic natural frequencies into the frequency range of interest for rotordynamics. For these circumstances, the conventional stiffness-damping-mass model cannot describe the force/motion relationship and hence, a general Laplace-transform transfer-function model is presented. Within the frequency range of interest for rotordynamics, the transfer-function models are found to be approximated by: (a) a lead-lag function for the direct term, and (b) a lag function for the cross-coupled term.;Comparisons of predictions with measured seal leakage are good by both viscous shear stress models. All predictions by both models are within 15% of the measured values.;For comparisons of measured versus predicted dynamic characteristics, there is no clear overall best model. Specifically, Hirs model yields the best predictions of radial impedance while, the New model yields the best predictions of the tangential impedance. Predictions of the radial impedance by both models are best at low seal exit Mach numbers and are worst at high seal exit Mach numbers. Generally, for the tangential impedance, the Hirs model underpredicts the magnitude, the New model overpredicts the magnitude, and both models predict the phase well. The stability is generally overpredicted by both models. |
